​
Data privacy statement

At Wonderflow, security is a fundamental value. This Privacy Statement describes how Wonderflow collects, uses, shares, or otherwise processes information within the context of the product Wonder.

​
Subprocessors

  • OpenAI (provider of Large Language Models) - US zone
  • Anthropic (provider of Large Language Models) - US zone
  • Amazon Web Services (provider of infrastructure) - EU zone
  • Google Cloud Platform (provider of infrastructure) - EU zone

​
International transfer of information

As previously described, Wonder leverages Large Language Models (LLMs) technologies for various machine-learning tasks.

In the current version, the LLMs in use are provided by OpenAI. As a consequence, there is data transfer to the US.

Wonderflow is planning and implementing a multi-LLM approach to Generative AI so that in the future, data transfer to the US will be removed when needed.

​
Data retention

For the OpenAI and Anthropic sub-processors, data retention is a maximum of 30 days for abuse and misuse monitoring purposes. Neither OpenAI nor Anthropic use data to train and improve Machine Learning Models.

For all other sub-processors and Wonderflow services, refer to the standard data retention of Wonderflow.

​
Data Accuracy

Generative AI and, in particular, Large Language Models (LLMs) are commonly affected by two limitations:

  • Hallucination: an LLM could give information that is not true
  • Out of date: an LLM could give information that is out of date

These two limitations commonly happen when the LLM solely bases its behavior on its knowledge.

Wonderflow removes these two limitations by avoiding the LLM using its internal knowledge to create the answer to the user questions.

In particular, Wonderflow uses sophisticated techniques (e.g., prompt engineering, LLM fine-tuning, in-context learning) that aim to anchor the LLM’s responses to information that is accurate, updated, and expansive enough to prevent hallucination.

Results are periodically tested and monitored via a rigorous quality control process.

​
Security

​
How we secure information processing

Information processing is secured in two main parts:

  • Wonderflow infrastructure and architecture: both are certified ISO27001, subject to yearly audits. Furthermore, they include multiple protection mechanisms to guarantee high security (for example, protection by account credentials with a strong password policy; VPN; malware detection; continuous integration and deployment via automatic testing with periodic security reviews, etc.). Please review the documentation on how Wonderflow handles security for further details on the protection mechanisms.
  • Wonderflow Trust Layer: a data processing component that identifies and anonymizes sensitive and personal information. This allows to prevent that sensitive and personal information is shared with sub-processors.

​
Changes to this Privacy Statement

Wonderflow will update this Privacy Statement from time to time to reflect changes in our practices, technologies, legal requirements, and other factors. We will update the “effective date” at the top if we do. If we make a material update, we may notify you before the update takes effect.

​
Contacting us

If you want to exercise your rights regarding your Personal Data or have questions about this Privacy Statement or our privacy practices, please email us at: security@wonderflow.ai.